Go App

A library to provide application level context, config reloading and log configuration. This is a companion to golangs context. It also tries to provide an extensible way for adding log hooks without requiring to download all of github.

This package is one of those tools you won't always need, but when you need it you'll know you need it.

Depends on

• logrus
• viper
• go-metrics

Includes

• tiny tracer
• modular initialization
• logging config through viper
• watching of configuration file, for online reconfiguration of loggers and modules
• watching of remote configuration, for online reconfiguration of loggers and modules

Config providers

By default the the application will look for config files in order of precedence:

• $HOME/.config/$APP_NAME
• /etc/$APP_NAME
• etc
• $CWD

You can customize those search paths through setting the environment variable: CONFIG_PATH eg. export CONFIG_PATH=/etc/my-app:etc

For the remote config providers you need to set a URL for the remote provider. You can optionally set a keyring, when present the remote configuration is expected to be encrypted with the public key of the gpg keyring.

To configure the url you need to set the CONFIG_REMOTE_URL environment variable:

export CONFIG_REMOTE_URL="etcd://localhost:2379/[app-name]/config.[type]"
export CONFIG_REMOTE_URL="consul://localhost:8500/[app-name]/config.[type]"

The extension of the file path is used to determine the content type for the key.

When you make a change to the config in the remote provider or in the local file the system will reload the loggers, and trigger the appropriate hook of registered modules.

Tracer

Using the tracer requires that you put a line a the top of a method:

var tracer = NewTracer("", nil, nil)

func TraceThis() {
  defer tracer.Trace()()

  // do work here
}

func FunctionWithUglyName() {
  defer tracer.Trace("PrettyName")()

  // do work here
}

You will then be able to get information about timings for methods. When you don't specify a key, the package will walk the stack to find out the method name you want to trace. If you think this is dirty, you can just pass a name to the trace method which will make you not incur that cost.

When used with the github.com/casualjim/middlewares package you can get a JSON document with the report from $baseurl/audit/metrics.

Modular initialization

Implements a very simple application context that does allows for modular initialization with a deterministic init order.

A module has a simple 4 phase lifecycle: Init, Start, Reload and Stop. You can enable or disable a feature in the config. This hooks into the watching infrastructure, so you can also enable or disable modules by just editing config or changing a remote value.

Each module is identified by a unique name, this defaults to its package name,

Usage

To use it, a package that serves as a module needs to export a method or variable that implements the Module interface.

package orders

import "github.com/casualjim/go-app"

var Module = app.MakeModule(
  app.Init(func(app app.Application) error {
    orders := new(ordersService)
    app.Set("ordersService", orders)
    orders.app = app
    return nil
  }),
  app.Reload(func(app app.Application) error {
    // you can reconfigure the services that belong to this module here
    return nil
  })
)

type Order struct {
  ID      int64
  Product int64
}

type odersService struct {
  app app.Application
}

func (o *ordersService) Create(o *Order) error {
  var db OrdersStore
  o.app.Get("ordersDb", &db)
  return db.Save(o)
}

In the main package you would then write a main function that could look like this:

func main() {
  app := app.New("")
  app.Add(orders.Module)

  if err := app.Init(); err != nil {
    app.Logger().Fatalln(err)
  }

  app.Logger().Infoln("application initialized, starting...")

  if err := app.Start(); err != nil {
    app.Logger().Fatalln(err)
  }

  app.Logger().Infoln("application initialized, starting...")
  // do a blocking operation here, like run a http server

  if err := app.Stop(); err != nil {
    app.Logger().Fatalln(err)
  }
}

Logger Configuration

The configuration can be expressed in JSON, YAML, TOML or HCL.

example:

logging {
  root {
    level = "debug"
    hooks = [
      { name = "journald" }
    ]

    child1 {
      level = "info"
      hooks = [
        {
          name = "file"
          path = "./app.log"
        },
        {
          name     = "syslog"
          network  = "udp"
          host     = "localhost:514"
          priority = "info"
        }
      ]
    }
  }

  alerts {
    level  = "error"
    writer = "stderr"
  }
}

or the more concise yaml:

logging:
  root:
    level: Debug
    hooks:
      - name: journald
    writer: stderr
    child1:
      level: Info
      hooks:
        - name: file
          path: ./app.log
        - name: syslog
          network: udp
          host: localhost:514
          priority: info
  alerts:
    level: error
    writer: stderr